Prostate cancer (PCa) is one of the most frequently occurring cancers and cause of cancer-related deaths in men [1]. According to the latest report from the American Cancer Society, in 2014, 233,000 new cases will be diagnosed, and 29,480 estimated deaths from PCa are predicted in the United States [1]. Although the five-year relative survival rate has increased over the past 25 years [2], PCa is still the leading cause of cancer death in older men [1]. There is an urgent need for a better understanding of the factors related to PCa development and prognosis.

Cyclooxygenase-2 (COX-2), an inducible enzyme, is overexpressed in PCa tissue [3] and plays a role in PCa cell growth [4]. Aberrant or high expression of COX-2 has been implicated in carcinogenesis and poorer prognosis [5, 6]. Therefore, it has been speculated that non-steroidal anti-inflammatory drugs (NSAIDs), which inhibit the COX pathway, may provide a strategy for mechanistically based PCa chemoprevention and therapy [3]. Indeed, a number of epidemiologic studies have investigated the association between NSAID use and the risk of PCa but showed conflicting results, with the majority finding null effects [7–22], some reporting statistically significant inverse associations [23–34] and a minority reporting a significantly elevated risk of prostate cancer in association with NSAID or non-aspirin NSAID (NA-NSAID) use [35–37]. In a study conducted by Mahmud et al. [38], which searched databases in June 2008, the association of the use of aspirin and other NSAIDs with PCa incidence was suggestive but not conclusive. Several observational studies published after this meta-analysis have shown contrasting results [19–22, 32–34, 36, 37], which has added new evidence to the previous research. On the other hand, since considerable evidence implies that aspirin use may reduce the risk of PCa, a better understanding of whether such therapy can influence disease outcome is important and necessary. To date, several studies have examined NSAID (aspirin) use and PCa-specific mortality, and these have reported conflicting findings [39–46].

Given widespread and long-term NSAID use, more knowledge is needed to clarify the drugs’ role in PCa incidence and mortality. Therefore, we performed a comprehensive review and provide a quantitative assessment of all relevant published studies to understand this issue better. A better understanding of the relationship may also highlight the importance of considering additional prevention methods in this area.

This meta-analysis, involving a total of 108,136 PCa cases from 39 observational studies, examined the association of the use of any NSAID with the risk of PCa incidence or mortality. The major finding of the present meta-analysis provides support for the mechanistic hypothesis that aspirin use (including long-term use of more than four years) is inversely related to the risk of PCa incidence. The association was slightly stronger for advanced PCa than for total PCa (OR = 0.92 for total PCa, OR = 0.81 for advanced PCa). A cumulative meta-analysis showed that a statistically significant effect of aspirin use on PCa incidence was consistently observed after publication of the 13th trial in 2006. More importantly, aspirin use demonstrated a 14% decrease in PCa-specific mortality for total PCa compared to non-use. In general, the pooled effects for any NSAIDs, NA-NSAIDs and COX-2 inhibitors demonstrated no adverse or beneficial effects on PCa development or PCa-specific survival, but the results were not consistent; all the effect estimates varied by geographic region.

There is a long-standing debate about the relationship between NSAID use and cancer. Several reviewed studies discussed the potential chemoprevention effects of NSAID use against tumors at various sites, including breast [57], lung [58] and brain [59], and melanoma [60]. This meta-analysis, with a much larger number of participants than previous meta-analyses [38, 61], adds to the previous findings by showing that aspirin use demonstrates a protective effect against PCa. With a larger sample size, increased statistical power could be obtained. In addition, we evaluated whether the association varied by COX-2 inhibitor, long-term aspirin use and study quality, and we investigated whether NSAID use influenced mortality from the disease, issues that were not discussed in the primary meta-analysis. Moreover, the present studies included an approved quality evaluation system; thus, it was more reliable in minimizing potential bias.

There are several proposed mechanisms by which these effects could be explained. Primary laboratory studies have pointed to COX-2 overexpression in PCa tissue [3], and it is well established that increased expression of COX-2 is related to tumor growth [4] and poorer prognosis [5]. NSAIDs inhibit PCa cell growth by reducing the synthesis of prostaglandins by COX enzymes [62], including inducing apoptosis [63], inhibiting cellular proliferation and angiogenesis [64]. Animal and laboratory studies have confirmed the preventive effect of NSAIDs on PCa repeatedly and consistently. In a review of 12 studies, the authors reported that, in all the included studies, all types of NSAIDs exhibited inhibitory effects on PCa development and progression.

In our study, use of aspirin was significantly inversely related to the risks of developing advanced PCa but not non-advanced PCa, which is difficult to explain. It may possibly be explained by detection bias if aspirin users were more likely to be screened and, therefore, to be diagnosed at an earlier stage. However, this effect may be real. Laboratory studies have suggested that advanced stage PCa may represent heterogeneous etiologies [65] and the mechanism by which aspirin acts on the initiation of poorly-differentiated disease may be distinct from its influence on the development of advanced disease [20]. There was also evidence that NSAIDs evoke tumor regression in gastrointestinal tumors [66], and may have similar effects against cancers of other sites including the prostate. Our findings with regard to the effects of aspirin on disease severity require confirmation in larger studies using unbiased diagnostic approaches.

We postulated that the use of aspirin reduced overall PCa risk. This could have significant implications with respect to the dose, frequency and duration of aspirin use. Unfortunately, most studies lacked information on dosage, frequency and duration of exposure. Although some studies have provided this information, they varied in each trial, resulting in invalid statistical analysis in these groups. In a study by Veitonmaki et al. [37] in 2013, a dose–effect relationship was found, indicating a significant inverse association (OR = 0.83) only among participants who used aspirin at the dosage of 37 to 1,300 defined daily dose (DDD). The other benefit was shown among those who took more than one aspirin pill per day [23, 29, 34], a low dose (≤75 mg/daily) [27, 32], larger doses (≥325 mg/daily) [26] or more than six tablets/week [20]; nevertheless, other studies found no evidence of a dose–effect [7, 10, 17, 19, 33, 35] or frequency–effect [14, 16, 17] relationship. In terms of the duration of aspirin use, we used the available data from eight studies [15, 17–20, 26, 27, 32] with durations of more than four years. Our pooled results demonstrated a negative trend of PCa risk with more than four years of aspirin use.

Considering the high incidence of PCa and the widespread use of aspirin in the general population, therefore, successful prevention could have a major public health impact. However, physicians should be aware of which subgroups of the population are at high risk for PCa, and the optimal dosage of aspirin, as well as its side effects, should be addressed. It seems that the use of aspirin in clinical prevention still has a long way to go, but, at least, the evidence implies that it is beneficial against the risk of PCa.

Nevertheless, several limitations of this study must be acknowledged. First, heterogeneity was a potential problem when interpreting the results of our analysis. In our meta-analysis, significant heterogeneity was observed in many of the analyses we conducted. Despite stratifying the data into subgroups based on type of outcome, study design, geographic region, quality of study and duration of aspirin use, some heterogeneity was still detected. In fact, it is no surprise given the differences of each study in the definition of drug exposure, information collection methods (for example, questionnaires or medical records), race, age and lifestyle factors of participants, sample size, duration of study follow-up and so on.

Second, as shown in Table 1 and Table 2, the number and content of the adjusted confounders differed among studies, which could have caused some imprecision in the effect estimates. The established risk factors for PCa are African-American ethnicity and family history of PCa [65]. Most studies adjusted for age and race using multivariate statistical models. Few studies adjusted for family history of PCa. However, researchers do not always make the same decisions concerning confounding factors. We did our best to minimize these confounding biases by choosing the most multivariable adjusted-effect estimates to analyze.

In addition, several sources of bias could have affected our observed associations. First, observational studies have inherent limitations, notably bias and unmeasured confounding. On the other hand, screening and surveillance biases are potentially a major source of systemic error in the reviewed studies. NSAID use may be associated with the socioeconomic status of patients. Thus, NSAID users might maintain healthier lifestyle habits than non-users, which could influence their risk for PCa. Moreover, NSAID users may have easier and more frequent access to preventive healthcare services, such as screening for prostate-specific antigen (PSA), which theoretically could lower the incidence of PCa.

At this stage, randomized controlled trials (RCTs) would be required to evaluate further the relationship between NSAIDs and PCa risk. However, RCTs for this topic could be extremely difficult or even impossible. First, PCa is rare and mainly occurs in older men; second, too many people in the general population take NSAIDs, especially aspirin, for various important medical reasons. This will render randomization and adherence to the protocol impossible. If one wants to conduct such a trial in younger subjects who do not take NSAIDS, investigators will run into the trouble of having enough end-points (events) given their young age.

Our current study, based on the available studies and updated data from a previous meta-analysis, supports the hypothesis that aspirin use (including long-term use) provides potential benefits in the reduction of PCa incidence and PCa-specific mortality. The inverse association was slightly stronger for advanced PCa than for total PCa, but the effect estimates varied by geographic region. Also unclear is the influence of dose and the frequency of aspirin use on PCa incidence and outcomes. Thus, caution needs to be exercised to ensure that the associated prevention benefits of aspirin outweigh the potential side effects (for example. gastrointestinal bleeding).